Question for the devs

Would it be possible to setup test systems that match the systems that are having problems and run with an ai observing the data flow to identify where and how these errors are happening? Then human coders can use that pinpointing to correct issues? I don't know what I'm talking about but I think it makes sense.